The need for (computing) speed
The advent of massively parallel computing gave scientists in many fields the ability to finally solve problems once considered intractable and run simulations of experiments that would be difficult, if not impossible, in the laboratory. Now, scientists in Japan plan to further increase the capabilities of this technology with a new supercomputer that promises to be faster than today’s 500 most powerful systems all put together.
The computer, which will be built by the collaborative efforts of NEC and Hitachi among others, is predicted to have a top speed of 10 petaflops, or 10^16 calculations per second (and cost ~$800 million to build); it would take roughly a billion desktop computers to do the same. IBM’s Blue Gene/L, the current titleholder, can only perform 1.4 x 10^14 calculations per second for comparison.
Most intriguing about this new supercomputer is its anticipated architecture. Today’s top systems (Blue Gene/L included) already use specially designed chips to enable faster memory access and other optimizations; however, the proposed design might feature an unprecedented hybridization of several different specialized processors in one machine. Such assembling of a variety of chips, produced to each perform a single calculation at incredible rates, has the potential to lend great bursts of speed to supercomputing’s very specific applications (e.g.: climate studies, protein folding simulations).
More pressing than questions of global warming and drug interactions, though, is… how can I get one of these?
No Comments »
No comments yet.
RSS feed for comments on this post.
| TrackBack URI
You can also bookmark
this on del.icio.us or check the cosmos